Allow None value in feature_scaling

doc/run_experiment.rst

@@ -530,19 +530,19 @@ Whether to scale features by their mean and/or their standard deviation. If you
 scale by mean, your data will automatically be converted to dense, so use
 caution when you have a very large dataset. Valid options are:
 
-none
+"none" or `None`
     Perform no feature scaling at all.
 
-with_std
+"with_std"
     Scale feature values by their standard deviation.
 
-with_mean
+"with_mean"
     Center features by subtracting their mean.
 
-both
+"both"
     Perform both centering and scaling.
 
-Defaults to none.
+The values are case insensitive. Defaults to "none".
 
 .. _featureset_names:
 

skll/config/__init__.py

@@ -720,9 +720,10 @@ def parse_config_file(
 
     # ensure that feature_scaling is specified only as one of the
     # four available choices
-    feature_scaling = config.get("Input", "feature_scaling")
+    feature_scaling = config.get("Input", "feature_scaling").lower()
     if feature_scaling not in VALID_FEATURE_SCALING_OPTIONS:
-        raise ValueError("Invalid value for feature_scaling parameter: " f"{feature_scaling}")
+        raise ValueError("Invalid value for feature_scaling parameter: "
+                         f"{feature_scaling}")
 
     suffix = config.get("Input", "suffix")
     label_col = config.get("Input", "label_col")

tests/configs/test_feature_scaling_none.template.cfg

@@ -0,0 +1,19 @@
+[General]
+experiment_name=test_class_map
+task=evaluate
+
+[Input]
+feature_hasher = true
+hasher_features = 100
+featuresets=[["test_class_map"]]
+learners=["LogisticRegression"]
+suffix=.jsonlines
+class_map={'dog': ['beagle', 'dachsund']}
+feature_scaling=None
+
+[Tuning]
+grid_search=False
+objectives=['accuracy']
+
+[Output]
+probability=false

tests/test_input.py

@@ -1848,6 +1848,32 @@ def test_config_parsing_set_wandb_values(self):
         self.assertEqual(wandb_credentials["wandb_entity"], "wandb_entity")
         self.assertEqual(wandb_credentials["wandb_project"], "wandb_project")
 
+    def test_config_parsing_feature_scaling_none(self):
+        """Test that feature scaling value of None is converted to a string"""
+        values_to_fill_dict = {
+            "experiment_name": "config_parsing",
+            "task": "evaluate",
+            "train_directory": train_dir,
+            "test_directory": test_dir,
+            "featuresets": "[['f1', 'f2', 'f3']]",
+            "fixed_parameters": '[{"estimator_names": '
+                                '["SVC", "LogisticRegression", "MultinomialNB"]}]',
+            "learners": "['VotingClassifier']",
+            "objectives": "['accuracy']",
+            "logs": output_dir,
+            "results": output_dir,
+        }
+
+        config_template_path = config_dir / "test_feature_scaling_none.template.cfg"
+
+        config_path = fill_in_config_options(
+            config_template_path, values_to_fill_dict, "default_value_save_votes"
+        )
+
+        configuration = parse_config_file(config_path)
+        feature_scaling = configuration[20]
+
+        self.assertEqual("none", feature_scaling)
     def test_config_parsing_set_wandb_missing_value(self):
         """Test that config parsing works as expected for when values are missing `wandb_credentials`."""
         values_to_fill_dict = {